A hybrid heuristic model for classification rule discovery

This dissertation studies hybrid heuristic models in the context of classification rule discovery. Nature inspired search algorithms such as Genetic Algorithms, Ant Colonies and Particle Swarm Optimization have been previously studied on data mining tasks, in particular, classification rule discovery. We extended this work by applying a hybrid model which combines GA, PSO and hill climbers, in same type of classification tasks. Such models have been tested and proved to be performing better than individual search algorithms, in various combinatorial optimization problems. Our research focused on studying the same kind of performance enhancements in classification rule discovery tasks. As a part of this dissertation, we developed a model for a hybrid heuristic based classifier and implemented different variations of it in Java. These algorithms have been benchmarked against the well known decision tree induction algorithm C4.5. Results have been compared in terms of prediction accuracy, speed and comprehensibility. Our results showed that, heuristic based classifiers can compete with C4.5 in terms of prediction accuracy on certain data sets and outperform C4.5 in general in terms of comprehensibility. C4.5 always outperformed heuristic based classifiers in terms of speed due to relative inefficiency inherent in heuristic based classification models. We also showed that hybridization of heuristics could bring improvements in terms of execution speed in comparison to plain heuristic based classifiers